Art of AI · Tech & Tools
Understand the tools.
Make something useful.
AI news in context, practical tutorials and tools worth understanding. Follow a guide, explore a workflow or find the right starting point for your next project.
News & context · 3 October 2026
From prompts
to workflows.
Comfy Agent, Ollama decision models and n8n Agents: three recent announcements, their practical implications and the questions worth asking before you adopt them.
Read the tools briefing →In this edition
- 1 October: Comfy Agent arrives in Comfy Cloud.
- 29 September: Ollama adds decision-model support.
- 25 September: n8n introduces its new Agents feature.
Source-based reporting with clear release dates, current availability and editorial context.
Penny’s critical guide · Tools and their consequences
The AI-o-philic Toolkit
Useful assistants, dangerous gods: choose a practical kit for writing, research, images and automation, then examine what happens when the helpful machine becomes indispensable.
Choose your tools. Keep your judgment. →Tutorials / Put an idea to work
Start small. Learn by making.
Local or online AI? Run a useful first comparison.
Use the same small brief in two tools, check the result and count the work you still have to do. Includes a sample input and a reusable test prompt.
From visual reference to repeatable image workflow
Turn an art-style reference into a clear brief. Vary one thing at a time and keep a record that makes your next image easier to direct.
01 / On your machine
Local AI tools
Local software gives you more control over where a model runs. Allow time for installation, model downloads and hardware checks; the application, model and any extensions may have different licences.
LM Studio
A desktop starting point for downloading and running language models. Useful for trying local conversations and working with documents through an app interface.
Before choosing: check your operating system, available memory and the requirements of the model you want to run.
Ollama
Run models and connect them to applications through a model-serving interface. A useful route when you want to integrate a model into your own tool or workflow.
Before choosing: Ollama also offers cloud models. Check the selected model and endpoint rather than assuming every request stays local.
ComfyUI
A node-based environment for building generative image, video and audio workflows. Useful when you want to see, adjust and reuse the individual steps of a process.
Before choosing: expect a learning curve around models, nodes and hardware. ComfyUI also provides a cloud option.
02 / In the browser
Online tools for creating & exploring
Hosted tools make it easier to begin without managing model installation. Compare account requirements, usage limits, export options and the terms for the material you upload or create.
ChatGPT
A conversational workspace for drafting, exploring ideas and working with supplied material. Start with a specific task, relevant context and the output format you need.
Good first use: turn an approved outline into a draft, then check the claims and revise it in your own voice.
Adobe Firefly
Generate and edit visual material in a browser-based creative environment. Useful for exploring compositions, image variations and edits within a defined creative brief.
Before choosing: features, credits and terms depend on the plan and selected model. Review the current details for your intended use.
Lexica
Browse AI imagery to explore subjects, composition and visual directions. Use a reference search to sharpen what you want to make before generating your own image.
Keep in mind: finding an image is not permission to reuse it. Check the rights and terms attached to any material you want to publish.
03 / Beyond a single prompt
Platforms, workflows & services
Hugging Face Spaces
Explore hosted machine-learning demos or create a small application for others to try. Spaces can help you understand an approach before deciding whether to build it into a larger system.
Check first: the individual Space’s publisher, visibility, model licence and handling of uploaded data. A public demo is not automatically a production service.
n8n
Connect tools into workflows and decide where AI should enter a process. Suitable tasks might include routing information, preparing drafts and moving reviewed results between systems.
Choose your responsibility: n8n offers managed cloud and self-hosted deployment. Self-hosting brings infrastructure, updates and maintenance into your scope.
These are selected starting points, not a ranking or an exhaustive catalogue. Features and plans change; each entry links to the provider’s own information.
Make the choice fit the work
Local, online or a combination?
Local or self-hosted
Consider this when control over deployment matters and you can support the hardware and software. Check memory, model size, setup time and who will maintain the system.
Managed online services
Consider these when a quick start and less infrastructure work are priorities. Check data settings, account permissions, subscription limits and whether you can export your work.
A hybrid workflow
Some tasks can stay on your machine while others use a hosted service. Draw the actual data path: which text, files and credentials move between which systems?
“Installed locally” does not mean “everything is offline.” Cloud models, remote APIs and extensions can still make network requests. Choose the tools and configuration that match the information you are handling.
Compare the complete cost: licences, usage, hardware, setup, review and maintenance.
04 / Keep good references close
Resource pages worth starting with
Explore Art of AI
- Art Style DatabaseBrowse visual styles, techniques and concepts. Compare examples and follow the database’s paginated index.
- Art Style DiaryEssays that connect visual language with technology, culture and everyday life.
- AI glossaryA reference for terminology you may encounter while exploring tools and models.
- AI blogs to readOur existing reading list: a further route into AI news and commentary. Check the date of individual articles.
Learn from original sources
- ChatGPT prompting guideOfficial guidance on explaining your goal, context, expected output and boundaries.
- Hugging Face LearnCourses and learning material for understanding and building with machine learning.
- The Met Collection APIExplore museum collection records and available images. Check each record’s public-domain status before reusing an image.
- The Metropolitan Museum of ArtStudy artworks in their museum context: a useful counterweight to searching only AI-generated references.
05 / Try one useful thing
A small test beats a long shortlist.
Pick a representative task and compare two approaches using the same input. Keep the original, the result and your corrections. That record will tell you more than a polished demonstration with an unknown brief.
- Name the deliverable. A reviewed paragraph, an image concept, a searchable document or a report draft.
- Choose suitable input. Use material you can share with the selected tool.
- Check the result. Look for factual errors, missing requirements, visual defects and awkward exports.
- Count the whole effort. Include setup, generation, checking and revision before deciding what helps.
Common questions
Do I need a powerful computer to begin?
Not for a browser-based service. Local model requirements vary considerably, so start with the application’s hardware guidance and a model that fits your machine. Image and video workflows can be particularly demanding.
Are free tools free to use for any project?
A free download, a free tier and permission for commercial use are separate things. Check the application, model, input material and service terms for the actual project. Also account for hardware, hosting and your time.
Where should I start if I mainly want to make AI art?
Begin with a visual goal and a few references from the Art Style Database. Try a hosted image tool for a quick first result, or explore ComfyUI if you want to learn how a configurable workflow is built.
Can Art of AI help us choose and implement tools?
Yes. Our business services cover tool selection, training, creative workflows, reporting and integration, with PRODAT’s software and engineering expertise where needed. Start with one real task and the systems you already use.
From exploration to application
Need a workflow
your team can use?
Art of AI can help you select tools, test a focused use case and plan the implementation. Explore our offers across business operations, creative production and engineering.
Curated by Art of AI. Links and descriptions reviewed 3 October 2026. Provider documentation is the source for current installation steps, features and plan details.